Regime-Dependent Performance of Fundamental Traffic-Flow Models across Motorcycle-Proportion Conditions in Urban Mixed Traffic
Year: 2026
Type: article
Abstract: Motorcycle-dominated urban traffic exhibits heterogeneous characteristics that may influence the suitability of fundamental traffic-flow models under different traffic composition conditions. This study evaluates the performance of the Greenshields, Greenberg, and Underwood models across different motorcycle-proportion regimes on divided urban roads of type 4/2T in Padang, Indonesia. Field observations were conducted on Jalan Prof. Dr. Hamka and Jalan Adinegoro, with traffic data collected during a 12-hour observation period and aggregated into hourly observations. The observed data were classified into three motorcycle-proportion regimes: Regime A (59% <= %SM < 65%), Regime B (65% <= %SM < 70%), and Regime C (%SM >= 70%). Speed–density relationships were fitted separately for each regime, and model performance was evaluated using the coefficient of determination (R2) as the primary criterion and root mean square error (RMSE) as a complementary measure. The results show that Underwood yielded the highest R2 values in Regimes A and B, at 0.9100 and 0.8657, respectively, while Greenberg yielded the highest R2 value in Regime C, at 0.9449. Greenberg also produced the lowest RMSE in Regime C, at 6.55 km/h. These findings indicate that the relative suitability of classical fundamental traffic-flow models varies across motorcycle-proportion regimes, supporting a composition-sensitive approach to model selection in motorcycle-dominated urban mixed traffic.
Institution: Andalas University
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7236206069
